Skip to main content
Research preview — APIs may change. GitHub

Geocomputation

Geocomputation is the processing of geospatial data to answer spatial questions: How far apart are these two locations? Is this point inside that boundary? Do these regions overlap? The answers seem simple, but computing them correctly — on a curved, non-Euclidean surface, with consistent precision and deterministic results — is harder than it appears.

The computational challenge

Spatial computation involves real complexity:
  • Reference systems — The Earth is not flat, and different coordinate systems represent it differently. Computing distance on a sphere vs. a projected plane vs. a geoid gives different results.
  • Computational geometry — Operations like polygon containment, intersection detection, and area calculation require robust algorithms that handle edge cases (self-intersecting polygons, antipodal points, degenerate geometries).
  • Precision and determinism — Floating-point arithmetic is not associative. The same computation can produce different results depending on operation order, hardware, or library version. For verifiable results, computation must be deterministic.
Professional geospatial software (QGIS, GDAL, PostGIS) has spent decades solving these problems. Astral builds on that foundation. (This is a pragmatic shift from our original plan to reimplement these algorithms in Solidity.)

Scope of geocomputation in Astral

Astral’s geocomputation capabilities span three areas, at different stages of maturity:
CapabilityStatusWhat it does
Location proof verificationShippingEvaluates location proof credibility
Geospatial operationsShippingDistance, containment, intersection, area, length
Geospatial AI/MLPlannedSpatial analysis, prediction, pattern detection

Verifiable geocomputation

Computation is only useful in adversarial contexts if you can trust the result. For most applications, that means running code on a server and trusting the operator. For applications where the spatial answer triggers real-world consequences that may carry an incentive to lie — a smart contract execution, a compliance determination, an autonomous agent decision — that trust model isn’t sufficient. Astral makes geocomputation verifiable through three approaches, at different stages of development: Trusted Execution Environments (v0 — shipped). Astral’s Compute engine runs inside a TEE (currently EigenCompute), which provides hardware-level isolation. The TEE guarantees that the attested code executed on the attested inputs, and that the signing key never left the enclave. This is the foundation of Astral Location Services, our hosted service. Zero-knowledge circuits (research). ZK proofs would allow verification of correct computation without any trusted hardware — a verifier could confirm the result was computed correctly without re-executing the computation or trusting a TEE manufacturer. This is an active research direction, not yet implemented. Smart contract verification (limited). Some spatial operations could theoretically run onchain, but gas costs and computational limitations make this impractical for most geospatial operations today.

Next: Astral Location Services

The hosted TEE service that runs verification and computation

See also: